In this module, the learners will be introduced to the course and its syllabus, setting the foundation for their learning journey. The course's introductory video will provide them with insights into the valuable skills and knowledge they can expect to gain throughout the duration of this course. Additionally, the syllabus reading will comprehensively outline essential course components, including course values, assessment criteria, grading system, schedule, details of live sessions, and a recommended reading list that will enhance the learner’s understanding of the course concepts. Moreover, this module offers the learners the opportunity to connect with fellow learners as they participate in a discussion prompt designed to facilitate introductions and exchanges within the course community.

In this module, students will gain foundational knowledge of parallel and multi-threaded programming, exploring the core principles that underlie the efficient utilisation of modern multi-core and many-core processors. Beginning with an overview of parallel programming concepts, this module covers different types of parallelism, including data parallelism, task parallelism, and pipeline parallelism. Students will also examine critical performance metrics like speedup, efficiency, and scalability, which help in evaluating the benefits and trade-offs of parallel approaches.

This module provides an in-depth exploration of multicore processor architectures, examining the design principles, performance considerations, and challenges involved in building efficient multicore systems. Students will study how multiple cores interact within a processor, focusing on memory hierarchies, caching mechanisms, and the role of parallelism in improving computational performance.

This module introduces students to the architectural principles of General-Purpose GPU (GPGPU) systems and the CUDA programming model. It explores the hardware components, including Streaming Multiprocessors (SMs), CUDA cores, and memory hierarchy, which form the foundation of GPU computing. The module also provides an overview of the CUDA programming model, emphasising its thread hierarchy, grid, and block organisation. By understanding these fundamental concepts, students will develop the ability to harness GPU architecture for high-performance parallel computing.

This module provides a comprehensive understanding of how CUDA executes programs on GPUs. It covers key concepts such as warps, warp scheduling, and resource partitioning, which are critical for understanding GPU hardware behaviour. The module delves into branch divergence and its impact on performance, offering strategies to minimise its effects. It also emphasises exposing parallelism effectively by leveraging CUDA’s hierarchical execution model. Students will learn how to design and optimise GPU programs by aligning with the underlying execution model to maximise efficiency and throughput.

The CUDA Memory Model & Streams and Concurrency module introduces students to the intricacies of memory hierarchy in CUDA, including global, shared, and local memory. It emphasises the importance of memory coalescing and efficient memory access patterns to optimise performance on GPUs. The module also covers CUDA streams, explaining how concurrent kernel execution and memory operations can be managed to enhance parallelism. By understanding these concepts, students will gain the ability to design GPU programs that maximise throughput and minimise latency.

This module explains in depth the difference between processes and threads and introduces multithreaded programming using pthreads library. Students are expected to learn about the various functions in pthreads library and implement those to solve real-world problems through a multithreaded approach. It also discusses precautions to take while developing an algorithm that uses multi-threading.

This module aims to introduce students to Distributed memory programming using the Message Passing Interface (MPI). Students will learn about the functions provided by the MPI library and their descriptions. It will enable students to develop parallel programming codes and also to convert a serial programmed code into a parallel code with the help of the MPI functions.

This module aims to introduce the shared memory programming model with the help of the OpenMP library. Students will gain exposure to the functions in the OpenMP library and methods to implement those in code to implement parallelism using shared memory. Students will explore the foundational concepts of OpenMP through videos and readings, starting with the basics of the library and progressing to more advanced topics such as reduction clauses, variable scoping, and mutual exclusion. Through worked examples like the Trapezoidal Rule and sorting functions, learners will understand how to parallelise loops, manage scheduling, and apply critical sections and locks for safe concurrent execution. The module also covers tasking in OpenMP and classic concurrency problems like producers and consumers.

This module will introduce the n-body problem in physics, examining its significance in simulating gravitational interactions among multiple particles. It will explore classical and modern algorithmic approaches to solving the n-body problem, followed by a discussion on their computational complexity. Emphasis will be placed on identifying opportunities for parallelisation, and students will analyse and implement efficient parallel solutions using the programming languages and parallel computing directives covered in the course.

This module focuses on hands-on implementations of the Sample Sort algorithm using OpenMP, Pthreads, MPI, and CUDA. Students will explore the strengths and limitations of each parallel programming model through practical coding exercises. The module includes performance benchmarking and comparative analysis of the implementations to highlight trade-offs in scalability, efficiency, and suitability for different architectures. By the end of the module, students will have a strong grasp of each API and be equipped to make informed decisions about the most appropriate tool for a given parallel computing task.

Final Comprehensive Examination